myisampack tries both ways to compress the column values. When starting to analyze the existing uncompressed data, it collects distinct column values up to a limit of 8KB. If there are more, it falls back to byte value compression for this column.
This means also that myisampack may use different algorithms for different columns. Besides a couple of other tricks, myisampack determines for every column if distinct column value compression or byte value compression is better. After that it tries to combine byte value compression trees of different columns into one or more code trees. This means that finally we may have less code trees than columns. Therefore the column information in the file header contains the number of the code tree used for each column. Some columns might not need a code tree at all. This happens for columns which have the same value in all records.
Copyright © 1997, 2014, Oracle and/or its affiliates. All rights reserved. Legal Notices